The structural arrangement and dynamics of the heteromeric GluK2/GluK5 kainate receptor as determined by smFRET

Abstract

Kainate receptors, which are glutamate activated excitatory neurotransmitter receptors, predominantly exist as heteromers of GluK2 and GluK5 subunits in the mammalian central nervous system. There are currently no structures of the full-length heteromeric kainate receptors. Here, we have used single molecule FRET to determine the specific arrangement of the GluK2 and GluK5 subunits within the dimer of dimers configuration in a full-length receptor. Additionally, we have also studied the dynamics and conformational heterogeneity of the amino-terminal and agonist-binding domain interfaces associated with the resting and desensitized states of the full-length heteromeric kainate receptor using FRET-based methods. The smFRET data are compared to similar experiments on the homomeric kainate receptor to provide insight into role of the differences in the conformational dynamics between the two to the functional differences.

Introduction

Ionotropic glutamate receptors are glutamate-activated tetrameric ion channels that are canonically known for their participation in excitatory synaptic transmission and are subdivided into classes α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA), kainate, and N-methyl-d-aspartate (NMDA). Ionotropic glutamate receptors have been shown to participate in the reception of excitatory synaptic signaling, the development of dendrites, and the regulation of pre-synaptic glutamate and GABA release (1–5). Deviation from normal function in ionotropic glutamate receptors has been linked to several pathological and disease states (5–7). Given their integral role in normal physiology and various disease states, it is critical to characterize the conformational states of these receptors associated with its functional states to aid future development of conformation-specific modulators for therapeutic applications (8).

There are now published X-Ray crystallographic structures and cryo-EM structures available for each class of ionotropic glutamate receptor (9–19). These structures show that these receptors are arranged as a dimer of dimers with a domain swapping configuration in the extracellular domains (Figure 1). Collectively these structures show that binding of agonist results in closing of the cleft at the agonist-binding domain, and the resulting conformational change propagates to the transmembrane regions leading to activation. The agonist-binding domain dimers remain coupled during the activated state and are believed to decouple upon initiation of desensitization, relieving the stress on the transmembrane segments induced by agonist-binding domain cleft closure (9–15,18–20). The majority of ionotropic glutamate receptor models available are of the AMPA and NMDA subtype with only three available for full-length homomeric GluK2 kainate receptors, one in the antagonist-bound form (15,21) and two in the agonist-bound form exhibiting desensitized state (12,15). The structural data available for the homomeric kainate receptor has laid a foundation for our understanding of the structural characteristics that give rise to the unique function of kainate receptors.

Figure 1.

Extracellular amino-terminal domain (ATD) and agonist-binding domain (ABD) represented as dimer of dimers in domain swapping configuration (A and D subunits form the dimer at the ABD and A and B at the ATD). Accessible cysteines modified to form cysless constructs for GluK5 and GluK2 subunits are shown as black spheres. Transmembrane domain (TMD) is represented as cartoon structure inside the membrane. A twin-strep tag is attached to the C-terminus of GluK5 subunit shown in green.

The heteromeric GluK2/GluK5 kainate receptor is known to be the most abundant kainate receptor expressed in the brain (22). Therefore, identifying the structural features that are unique to the GluK2/GluK5 heteromer is crucial to design kainate receptor specific compounds with therapeutic potential. Yet, the only structural models currently available for the heteromeric kainate receptor are of the isolated amino-terminal domains (23). Functionally the GluK2/GluK5 heteromer has shown higher glutamate sensitivity (EC_{50 GluK2/GluK5} = 6 μM) compared to the homomeric receptor (EC_{50 GluK2/GluK2} = 186 μM) (24–28) and a faster rate of desensitization (τ_{des GluK2/K5} = ~3 ms) compared to the homomeric receptor (τ_{des GluK2} = 5.8 ms) in the presence of sodium ions (29). These differences in function leave no question that there are structural differences in the heteromeric kainate receptors that remain largely unresolved. What is known is that the GluK2/GluK5 heteromer that localize to the plasma membrane have a subunit stoichiometry of 2:2 and that the GluK2 amino-terminal domains likely mediate the amino-terminal domain dimer-dimer interface (23,30). Additionally, MD simulations on the homology model of isolated agonist-binding domain dimer show that the heteromeric receptor has more decoupled dimer interface relative to the homomeric receptors (29). However, these simulations were performed using the agonist bound isolated domain and changes in the context of the full-length receptor as well as the conformation in the resting apo state are not known.

FRET acts as a molecular ruler providing distances between the donor and acceptor fluorophores attached to specific sites of a molecule, and when used at the single molecule level allows for investigating the conformational landscape and energetics of functionally-significant dynamics within molecules (31–38). For the single molecule FRET (smFRET) measurements, membrane preparations of cells expressing the protein used with minimal purification, thus providing insight into the proteins in a near native conformation (31–38). Here, we have used smFRET to address the specific arrangement of the receptor, and to understand the conformational landscape of the heteromeric receptor, specifically the conformational landscape across the dimer-dimer interface at the amino-terminal domain, the dimer interface at the agonist-binding domain and at the transmembrane segments. These sites were chosen as they are known to play important roles in activation and desensitization in the homomeric kainate receptor and/or the closely related AMPA receptor (39–43). Additionally, the sites at the amino-terminal domain and agonist-binding domain are equivalent to those that we used to study the homomeric receptors (44), thus a direct comparison of differences between the conformational landscape of the heteromeric and homomeric receptors can be made and correlated to differences in function.

Results and Discussion

For smFRET measurements we modified the GluK2 and GluK5 receptors to remove accessible cysteines as shown in Figure 1 (modified constructs are referred to as GluK2* and GluK5*). Additionally, we have introduced a twin strep tag on the background of GluK5 constructs. Since, GluK5 subunits do not express as homomeric receptors (30,45–48), performing the in situ pull down using streptavidin on membrane preparations of HEK-293 cells co-expressing GluK2 and GluK5 subunits allows for the specific attachment of GluK2/GluK5 receptors and excludes the attachment of GluK2 homomeric receptors.

The sites for fluorophore attachment were selected based on two criteria. Firstly, the sites must reflect large-scale conformational rearrangements as interpreted from currently available antagonist- and agonist-bound kainate and AMPA receptor structures. Second, the geometry of the labeling sites within the receptor should be such that the different distances within and across the dimer exhibit distinct and well separated FRET efficiencies (38). Table 1 shows differences in FRET efficiencies calculated based on alpha carbon distances within and across the dimer using homology model generated based on known structures of AMPA and kainate receptors. These calculated FRET efficiencies provide a reference point as they are based on backbone distances and not distances between fluorophores.

Table 1:

Summary of smFRET efficiencies and distances calculated

		APO STATE				GLUTAMATE BOUND STATE
SITES				3KG2 (AMPA)	5KUH (Kainate)			5KUF (Kainate)	5VHZ (AMPA)	4U4F (AMPA)	5WEO (AMPA)
		Observed FRET Efficiency (Calculated Distance in Å)	Gaussian fit	Expected FRET efficiency (Distance in Å for sites)		Observed FRET efficiency (Calculated Distance in Å)	Gaussian fit	Expected FRET efficiency (Distance in Å for sites)
GluK2*−266 - GluK2*−266		0.72 (43) 0.83 (39)	0.72 ±0.03 0.84 ± 0.004	0.70 (44)	0.62 (47)	-	-	-	-	-	-
GluK2*−479 - GluK5*−471		0.80 (40) 0.90 (35)	0.80 ± 0.01 0.90 ± 0.01	0.83 (39)	0.87 (37)	0.56 (49) 0.69 (45) 0.83 (39) 0.94 (32)	0.56 ± 0.01 0.68 ± 0.01 0.83 ± 0.01 0.93 ± 0.01	0.33 (57)	0.62 (47)	0.83 (39)	0.89 (36)
GluK2*−479 - GluK5*−471		Not observed		0.23 (62)	0.13 (70)	Not observed		0.20 (64)	0.22 (63)	0.25 (61)	0.20 (64)
GluK5*−471 - GluK5*−471		Not observed		0.05 (82)	0.04 (87)	Not observed		0.02 (97)	0.05 (85)	0.08 (77)	0.05 (83)
GluK2*−479 - GluK2*−479		Not observed		0.15 (68)	0.14 (69)	Not observed		0.11 (72)	0.12 (71)	0.15 (68)	0.24 (62)
GluK2*−523 - GluK2*−523		0.96 (30) 0.83 (39)	0.96 ± 0.01 0.83 ± 0.01	0.91 (34)	-	0.85 (38) 0.94 (32)	0.83 ± 0.01 0.94 ± 0.01	0.85 (38)	0.87 (37)	0.90 (35)	0.71 (44)
GluK5*−515 - GluK5*−515		0.94 (32) 0.81 (40)	0.94 ± 0.01 0.81 ± 0.01	0.91 (34)	-	0.84 (38) 0.93 (33)	0.85 ± 0.01 0.95 ± 0.01	0.85 (38)	0.90 (35)	0.91 (34)	0.87 (37)

Arrangement of GluK2 and GluK5 subunit with the kainate receptor.

Given that the kainate receptor subunits are arranged as dimer of dimers, the GluK2 and GluK5 subunits can be assembled in three possible configurations (Figure 2). To determine the configuration(s) that the receptor occupies using smFRET, we measured distances between sites 266 and 266 in the two GluK2 subunits, and equivalent sites 272 and 272 in the two GluK5 subunits, within the GluK2/GluK5 heteromeric receptors. These sites were chosen based on homology models of the GluK2/GluK5 receptors generated, which show that the distances between these sites can be used to clearly differentiate between the possible configurations (Figure 3 A). The functionality of GluK2*−266/GluK5* and of GluK2*/GluK5*−272 subunits labeled with Alexa 555 and Alexa 647 fluorophores were established using whole cell current recordings (Figure 4).

Figure 2.

Possible configurations for amino terminal domains of GluK2 and GluK5 subunits in GluK2/GluK5 heterotetramer forming dimer of dimers with distance calculated for K2-K2 subunits and K5-K5 subunits based on homology model for each possible configuration and smFRET experiment.

Figure 3.

(A) Full-length structure of the apo state GluK2/GluK5 heteromer (homology model made from PDB 3KG2) with the GluK2 subunits shown in blue and the GluK5 subunits shown in green. Alpha carbon sites at GluK2*−266 are shown as red spheres (B-C) smFRET data for GluK2*−266 sites; with two representative smFRET efficiency traces for individual molecules shown in panel B and cumulative smFRET efficiency traces with observed data (grey) overlaid on denoised data (red) are shown in panel C. Gaussian fits shown in black, blue and green, represents the smFRET efficiency states.

Figure 4.

FRET construct characterization. Representative whole-cell recording for GluK2*−479/GluK5*−471, GluK2*/GluK5*−515, GluK2*−523/GluK5*, GluK2*−266/GluK5* and GluK2*/GluK5*−272.

smFRET traces for constructs GluK2*−266/GluK5* exhibiting a single donor and single acceptor photobleaching step with anticorrelation upon acceptor bleaching were used for generating the smFRET efficiency traces for individual molecules (representative trace shown in Figure 3 and example trace shown in supplemental Figure 1). The fractional occurrence in the smFRET efficiency traces were then used to generate the smFRET efficiency histogram. The number of states that best describes the dataset was determined using Step Transition and State Identification (STaSI)(49). STaSI helps in the identification of transitions within traces by using a t-test and determines the number of states represented in a dataset with a minimum description length algorithm (49), ensuring that the states determined are statistically significant. The number of STaSI-determined states and their relative occurrences were used as guiding parameters during the process of fitting the Gaussian curves to the observed histograms, and this produced fits and standard error for the fits are shown in Table 1. The cumulative histogram from smFRET efficiency traces from 28 molecules, for GluK2*−266/GluK5* is shown in (Figure 3C). The donor and acceptor traces were denoised using wavelet based denoising and the denoised FRET histograms generated from these are shown overlaid on the observed histogram (Figure 3 C). The denoised smFRET histogram shows primarily two states with a FRET efficiency of 0.72 and 0.83, corresponding to distances of 43 Å and 39 Å. Gaussian fits of the observed data corelate well showing efficiencies of 0.72 ±0.03 and 0.84 ±0.01. These distances are close to the alpha carbon distance between the 266 and 266 residues (values are shown in Table 1) as predicated when placing the GluK2 subunits proximal to each other (Figure 2A). Additionally, given that this distance is significantly shorter than the distance of 67 Å expected in the configuration placing two GluK2 subunits within the dimer as seen in GluK2 homomer (PDB 5KUH), it can be concluded that the configuration with GluK2 (Figure 2B) within the peripheral position the dimer does not exist in the heteromer.

smFRET experiments for the GluK2*/GluK5*−272 receptors on the other hand showed no molecules with significant smFRET efficiency traces. 148 molecules probed exhibited one or multiple donor steps or one or multiple acceptor steps, but no molecules showed FRET between the donor and the acceptor. Based on this observation it can be concluded that no significant fraction of the GluK2/GluK5 heteromeric receptors exist in the configuration placing the GluK5 subunits proximal to each other (Figure 2C) and within a dimer (Figure 2B). Therefore, the full length heteromeric receptor expressed in HEK-293 cells exists primarily in the configuration shown in Figure 2A.

Conformational landscape of the amino terminal domain and comparison with homomeric kainate receptors.

In order to correlate these FRET efficiency based states to possible structural changes we modeled the heteromeric kainate receptor into the known kainate and AMPA receptor structures. The alpha carbon distances based on these models for the different structures are listed in the Table 1 along with the calculated FRET efficiencies. These act as reference points for correlations with our smFRET data (with the focus being changes in distances and hence packing). The model of the heteromeric kainate receptors generated from the antagonist bound structure of homomeric kainate receptor shows a slightly looser packing at the amino terminal domain relative the antagonist bound structure of AMPA receptor, showing a 3 Å increase in distance between these two structures at site 266 across the dimer-dimer interface. A similar distance change (4 Å) is seen between the two states at this site in the smFRET efficiency states for the heteromeric receptor suggesting that it is possible for the amino terminal domain to exist in two possible configurations with small variations in degree of coupling, and such differing degrees of packing can be accounted for with these known structures.

While, the smFRET data of the homomeric GluK2 receptors showed three states of FRET efficiencies 0.69, 0.83 and 0.92 (44) compared to the two observed here for the heteromeric receptors, the most probable state of 0.83 is identical in both cases. Thus, suggesting that the two receptors have a similar conformational arrangement of the dimer-dimer interface at the amino-terminal domain for the primary conformational state. The similarity in primary conformation and lack of large-scale differences at the amino-terminal domain between the homomeric and heteromeric receptors would be consistent with the previous biochemical studies suggesting that amino-terminal domain plays a major role in assembly (23,50).

Conformational landscape of the agonist-binding domain.

To study the conformational dynamics at the dimer interface in the agonist-binding domain, we introduced cysteines at site 479 on GluK2* and 471 on GluK5* (Figure 5 A). These sites are ideal as the distance between these residues are distinct within the dimer, relative to that across the dimers. smFRET traces showing a single donor and single acceptor photobleaching step with anticorrelation between the two were used to generate the FRET efficiency traces, and 26–28 molecules were combined to generate the cumulative smFRET efficiency histograms. The denoised smFRET histograms in the apo state show two states with efficiencies of 0.80 and 0.90 (Figure 5C), corresponding to distances of 40 Å and 35 Å. Gaussian fits of the observed data corelate well showing efficiencies of 0.80 ±0.003 and 0.90 ±0.001. These distances and changes in distances again correspond well with the known structures of antagonist bound forms of kainate and AMPA receptors (Table 1). Additionally, the smFRET data at the agonist binding domain are in agreement with the data at the amino terminal domain showing that there are two possible configurations in apo state, with slight variations in the degrees of decoupling at both domains, but that in both domains the tightly coupled state is more favored.

Figure 5.

(A) Full-length structure of the apo state GluK2/GluK5 heteromer (homology model made from PDB 3KG2) with alpha carbon sites at GluK2*−479 (magenta spheres) and GluK5*−471 (red spheres). (B-C) smFRET data for GluK2*−479 and GluK5*−471 sites at the apo state (left panel) and the desensitized state (right panel). (B) Two representative smFRET efficiency traces for individual molecules. (C) Cumulative smFRET efficiency traces with observed data (grey) overlaid on denoised data (red). Gaussian fits shown in black, blue and green, represents the smFRET efficiency states.

The denoised smFRET histograms for the glutamate-bound state, on the other hand, shows four states with FRET efficiencies of 0.56, 0.69, 0.83 and 0.94 (Figure 5C). Under these conditions the receptor is primarily expected to be in the desensitized state. Gaussian fits of the observed data corelate well showing efficiencies of 0.56 ±0.01, 0.68 ±0.01, 0.83 ±0.01 and 0.93 ±0.01. The distances associated with each if these FRET efficiencies are listed in Table 1. These FRET efficiencies span the range of distances from 49 to 32 Å (difference of 17 Å.) The distance range for the models based on known agonist bound state structures are 57 to 36 Å (difference of 21 Å) (Table 1). The range of decoupling seen in the smFRET measurements is hence closely related to the range seen in the known structures. However, it is interesting to note that the structure of homomeric kainate receptor structure thought to be in the desensitized state (PDB 5KUF) shows the largest decoupling (15). The smFRET data for the heteromeric receptor, on the other hand, shows that the largest decoupled state in fact represents the lowest occupancy state, and the highest occupancy state of FRET efficiency shows intermediate decoupling which would correspond to the structure showing only slight decoupling such as that seen for the agonist bound form of AMPA receptors PDB:4U4F (13). These results suggest that complete decoupling of the dimer interface is not necessary for desensitization in the kainate receptors.

Comparison of the conformational landscape of heteromeric to homomeric receptors at the agonist-binding domain dimer interface.

While, the FRET efficiencies of 0.56, 0.69, 0.83 and 0.94 in the glutamate-bound state for the heteromeric receptors are similar to the FRET efficiencies of 0.54, 0.69, 0.82 and 0.93 seen in the homomeric GluK2 receptors, the fractional occupancy is higher for the more decoupled state in the heteromeric receptors relative to the homomeric receptors. This shift towards the decoupled states, indicative of decreased ABD dimer stability, would be consistent with the faster desensitization rates observed in the heteromeric receptors relative to the homomeric receptors (τ_{des GluK2/K5} = ~3 ms, τ_{des GluK2} = 5.8 ms) in the presence of sodium ions (29). The smFRET results are also consistent in trends with MD simulations which show a decoupling of the dimer interface in the heteromeric receptor relative to homomeric receptor, however the shifts were small showing a shift of 1 Å towards a more decoupled state (29). The lack of large changes in the MD simulations could be due to the shorter time scales of the simulations or due to the fact that they were performed on the isolated agonist-binding domain.

Conformational landscape at the transmembrane segments.

In order to study the conformational dynamics at the transmembrane segments we introduced cysteines at site 523 on GluK2* and 515 on GluK5* (Figure 6, 7). These sites are positioned at the top of the first transmembrane segment, which makes them ideal to measure the distance across the pore axis at complementary sites. smFRET traces showing a single donor and single acceptor photobleaching step with anticorrelation between the two were used to generate the FRET efficiency traces, and 27–31 molecules were combined to generate the cumulative smFRET efficiency histograms. The denoised smFRET histograms for GluK2*−523 in the apo state show two states with efficiencies of 0.83 and 0.96 (Figure 6C) and correspond to distances of 39 and 30 Å. Gaussian fits of the observed data corelate well showing efficiencies of 0.83 ±0.005 and 0.96 ±0.002. These distances are close to the alpha carbon distance of residue 523 – 523 in the two GluK2 subunits obtained from the apo state homology model with the channel being in a closed state (Table 1). The smFRET histograms for GluK2*−523 in the glutamate-bound state show two peaks with efficiencies of 0.85 and 0.94, corresponding to distances of 38 and 32 Å. These efficiencies are similar to what is found under apo conditions at this site suggesting similar packing in the resting and desensitized closed states; however, there is a significant shift in occupancy toward the high FRET state in the glutamate bound form, suggesting the more tightly packed transmembrane conformation is favored in the desensitized state. The tighter packing observed under desensitized conditions differs from distances found in the homology model (Table 1). These differences likely result from additional stability at the transmembrane domains due to the retention of lipids around the transmembrane regions in measurements using the SiMPull(59) method.

Figure 6.

(A) Full-length structure of the apo state GluK2/GluK5 heteromer (homology model made from PDB 3KG2) with alpha carbon sites at GluK2*−523 (red spheres). (B-C) smFRET data for GluK5*−523 sites at the apo condition (left panel) and the desensitized state (right panel). (B) Two representative smFRET efficiency traces for individual molecules. (C) Cumulative smFRET efficiency traces with observed data (grey) overlaid on denoised data (red). Gaussian fits shown in black, blue and green, represents the smFRET efficiency states.

Figure 7.

(A) Full-length structure of apo state GluK2/GluK5 heteromer (homology model made from PDB 3KG2) with alpha carbon sites at GluK5*−515 (red spheres). (B-C) smFRET data for GluK5*−515 sites at apo condition (left panel) and desensitized state (right panel). (B) Two representative smFRET efficiency traces for individual molecules. (C) Cumulative smFRET efficiency traces with observed data (grey) overlaid on denoised data (red). Gaussian fits shown in black, blue and green, represents the smFRET efficiency states.

The smFRET histograms for GluK5*−515 in the apo state show two states with efficiencies of 0.81 and 0.94 and correspond to distances of 40 Å and 32 Å, and in the glutamate-bound state show two peaks with efficiencies of 0.84 and 0.93, corresponding to distances of 39 Å and 33 Å (Figure 7). These smFRET data from site GluK5*−515 are similar to site GluK2*−523 in the glutamate-bound state showing that at this site the protein exhibits a four-fold symmetry in both the apo and glutamate-bound state.

Conclusion

There are a plethora of X-ray crystallography and cryo-EM structural models available for the AMPA and NMDA classes of iGluR including both isolated amino-terminal and agonist-binding domains, and also the full-length receptors. However, only three structural models have been produced for the full-length homomeric GluK2 receptor type. Currently, there are no structures of the full-length GluK2/GluK5 heteromeric receptor. Using smFRET investigations, we show that GluK2/GluK5 heteromeric receptor assemble in one configuration with the GluK2 sites occupying proximal positions across the dimer-dimer interface at the amino-terminal domains of the receptor. Additionally, we show that the spread of conformational states is not significantly different between the homomeric and heteromeric receptors at the dimer-dimer interface at the amino-terminal domain suggesting that the primary role of this domain is in assembly. The agonist-binding domain, on the other hand, shows more decoupling and a higher occupancy of the decoupled state at the dimer interface in both the apo and glutamate-bound states of the heteromeric receptors relative to what is observed in the homomeric receptors. Prior studies have shown that the decoupling of the dimer interface at the agonist-binding domain is the primary conformational change driving desensitization. Therefore, the increase in decoupling at this interface in the agonist binding domain of the heteromeric receptor results in easier transition into the desensitized state and ties back to the functional studies that show a faster desensitization rate in the heteromeric receptor relative to the homomeric receptors. The smFRET studies also show that the GluK2/GluK5 heteromeric receptors loses its two-fold symmetry seen in the extracellular domain and exhibits four-fold symmetry at the start of the first transmembrane segment in both the apo and the glutamate-bound forms of the receptors, similar to the other known structures of kainate and AMPA receptors.

Methods

Homology Modeling

Six homology structures were built for heteromeric GluK2/GluK5, based on homomeric GluK2 antagonist bound form (PDB 5KUH), GluK2 agonist-bound form exhibiting desensitized state (PDB 5KUF), antagonist bound form of AMPA receptor (PBD:3KG2), and agonist bound forms of AMPA receptor (PDB: 5WEO, 5VHZ, 4U4F). For 5VHZ, 4U4F the agonists were also replaced by glutamate. The homology structures were built using MODELLER software (61).

Generation of FRET constructs

The R. norvegicus GluK2 construct used previously in cryo-EM (12,15) and smFRET (44) experiments was used and retained the native glutamine at site 590. The GluK2 coding sequence was PCR amplified and inserted into pcDNA3.1. The background GluK2 FRET construct was created by mutating the non-disulfide bonded cysteines C91, C199 and C432 to serines. From this background five constructs were created. One with S266 mutated to cysteine, one with A479 mutated to cysteine, one with S523 mutated to cysteine, one with both A479C and D776K mutations, and one with both S523C and D776K mutations.

The R. Norvegicus construct containing GluK5 was kindly provided by Geoffrey Swanson, PhD. The GluK5 coding sequence was PCR amplified and inserted into pcDNA3.1. The background GluK5 FRET construct was created by mutating the cysteines C14, C88 and C270 to serines. From this background five constructs were created. One with S265 mutated to cysteine, one with A471 mutated to cysteine, one with S515, one with both A471C and D776K mutations and one with both S515C and D776K mutations.

Electrophysiology

HEK 293T cells at 30% confluency were transfected using lipofectamine 2000 (Invitrogen) with GluK2 and GluK5, cotransfected with GFP at a microgram ratio of 1:4:0.5. Whole cell patch clamp recordings were performed 24–48 h after transfection, using firepolished borosilicate glass (Sutter instruments) pipettes with 3–5 mΩ resistance, filled with internal solution: 110 mM CsF, 30 mM CsCl, 4 mM NaCl, 0.5 mM CaCl2, 10 mM HEPES, and 5 mM EGTA (adjusted to pH 7.4 with CsOH). The external solutions contained 150 mM NaCl, 2.8 mM KCl, 1 CaCl2 and 10 mM HEPES (adjusted to pH 7.4 with NaOH). The glutamate (10 mM) was applied to cells using a stepper motor system (SF-77B; Warner Instruments) with Triple barrel tubing. Recordings were performed using an Axopatch 200B amplifier (Molecular Devices) at −60 mV hold potential, acquired at 10 kHz using pCLAMP10 software (Molecular Devices) and filtered online at 5 kHz.

smFRET sample preparation

HEK293T cells grown on 10cm plates were checked for approximately 50% confluency. The cells were then transfected with 20 μg DNA per 10 cm plate following JetPrime protocol. DMEM media was changed after four hours of transfection and was left for overnight expression. Next day, cells were collected from two transfected 10 cm plates and were washed with 3mL of extracellular buffer (ECB). The sample was then wrapped in foil and was labeled with 400 nM of Alexa 555 maleimide (ThermoFisher), a donor fluorophore and 400 nM of Alexa 647 maleimide (ThermoFisher), an acceptor fluorophore, in 3mL ECB at room temperature for 1hr. The labelled cells were washed with 3mL ECB and were resuspended in 2mL of solubilization buffer by nutating at 4°C for one hour. Solubilization buffer consists of phosphate-buffered saline, 1% lauryl maltose neopentyl glycol (Anatrace), 2 mM cholesteryl hydrogen succinate (MP Biomedicals), and 1/4 protease inhibitor tablet (Pierce). The nutated sample was then transferred to an ultracentrifuge tube and was spun for one hour at 44000 rpm at 4°C using a TLA 100.3 rotor for filtering unsolubilized debris. Supernatant thus collected were used as smFRET samples and were kept on ice until they were used.

smFRET slides preparation

Microscope glass slides (20 × 20 mm) were cleaned in a solution of Liquinox phosphate-free detergent (Alconox Inc.) via bath sonication followed by washing with solution consisting 4.3% NH₄OH and 4.3% H₂O₂. Slides were then washed with purified water, dried with nitrogen gas and placed in metal slide holder. Plasma cleaning of the slides was done using Harrick Plasma PDC-32G Plasma Cleaner and then the slides were treated with Vectabond (Vector Laboratories, CA) for aminosilanization and stored under vacuum. Clean silicone templates (Grace bio-Labs) which were bath sonicated, and methanol treated were dried using nitrogen flow and were placed at the center of the slides. The slides were then treated with 50μl of PEG solution (0.25% w/w biotinylated PEG, 25% w/w mPEG-succinimidyl carbonate, 0.1M NaHCO₃) and incubated in a dark moist environment overnight. On the day of the experiment, after cleaning the slides with purified water and dried with nitrogen, slides were applied with short chain PEG solution (25mM short-chain 333 Da MS(PEG)₄ Methyl-PEG-NHS-Ester Reagent, 0.1 M NaHCO₃) and were incubated at room temperature for two to three hours. Then, the slides were washed with water, dried with nitrogen and silicone templates were removed followed by applying Hybridwell chambers and press-fit tubing connectors (Grace bio-Labs). 36 μl of streptavidin solution (0.2 mg/mL streptavidin, 1×smFRET imaging buffer (1mM DDM (n-dodecyl-β-D-maltoside), 0.2mM CHS (cholesteryl hydrogen succinate), 1×PBS) was applied to the chamber, incubated for 10 minutes and was washed with 1×PBS. The smFRET sample was then applied to the slide according to the in situ immuno-precipitation SiMPull method(59) and incubated at 4°C for 20 minutes followed by washing of slides two times with 60 μl ROXS (reactive oxygen species) scavenging solution (3.3% w/w glucose, 0.1mg/mL pyranose oxidase, 0.01 mg.mL catalase, 1mM ascorbic acid, 1mM methyl viologen and/or 1mM glutamate) (Sigma-Aldrich). Finally, the slides were ready for imaging.

smFRET data collection

A PicoQuant MicroTime 200 Fluorescence Lifetime Microscope, a kind of confocal microscope, was used for acquiring the smFRET data with pulsed interleaved excitation (PIE) set at 80 MHz. The fluorophores were excited by using 532 nm (LDH-D-TA-530; Picoquant) and 637 nm (LDH-D-C-640; Picoquant) lasers simultaneously. The sample slide positioned on a scanning x-y-z piezo stage (P-733.2CD; Physik Instrumente) was observed through an oil immersed 100X objective lens (100× 1.4 NA; Olympus). Two SPAD photodiodes (SPCM CD3516H; Excelitas technologies) collected photons emitted from the sample passing through objective to the emission filters 550 nm (FF01–582/64; AHF/Semrock) and 650 nm (2XH690/70;AHF) to the photodiodes. These emission filters helps us to visualize the donor and the acceptor channel.

smFRET data analysis

Molecules exhibiting a single donor and single acceptor photobleaching step with anticorrelation upon acceptor bleaching were used for analysis. Intensities for donor and acceptor were then used for the calculation of FRET efficiencies. MATLAB (MathWorks) was used for denoising donor and acceptor traces using wavelet based denoising. Origin (OriginLab Corp) was used for creating smFRET histograms and traces. The number of states which best describes the efficiencies obtained for each dataset was determined using Step Transition and State Identification (STaSI)(49) and Gaussian fitting. The total number of molecules used for the analysis of each conditions are, 28 molecules for GluK2*−266 - GluK2*−266, 26–28 molecules for GluK2*−479 - GluK5*−471, and 27–31 molecules for GluK2*−523 - GluK2*−523 and GluK5*−515 -GluK5*−515.